Data Centers: Build Them or Fall Behind
Rejecting Luddism — Capturing Technology’s Gains for Washington Families
Fellow Washingtonians,
Data centers are not optional infrastructure in the AI era. They are the factories of the next industrial wave. Regions that treat them as enemies will not stop the technology. They will simply watch the investment, the jobs, the tax base, and the downstream prosperity move elsewhere.
Seattle’s 2026 one-year moratorium on large data centers (above roughly 20 MW) is a clear local example of symbolic restriction that does nothing to reduce global demand for compute. It only raises the odds that the next wave of capacity — and the economic activity that follows it — lands in other states or countries. That is how places fall behind.
Legitimate concerns exist: grid capacity, rate impacts on ordinary customers, water use, noise, and land-use conflicts. Those deserve targeted tools — full marginal-cost rates for large loads, water and noise standards, heat recovery where district systems already exist. None of those tools requires a freeze on applications.
The technology will advance whether we like it or not. The only open question is whether Washington captures the gains or exports them.
Power and Waste Heat in Context
Large facilities are measured in hundreds of megawatts. All electricity consumed eventually becomes heat that must be rejected. The relevant comparison is not neighborhood folklore — it is planetary energy flow.
Gas-fired plant + 500 MW data center
| Component | Power (MW) | Notes |
|---|---|---|
| Fuel input at plant | 1,000 – 1,250 | Natural gas |
| Waste heat at plant | 500 – 750 | Cooling towers / exhaust |
| Electricity to data center | 500 | Delivered output |
| Waste heat at data center | 500 | Essentially all electricity becomes heat |
| Total waste heat to environment | 1,000 – 1,250 | Combined |
That is approximately 1/97,600,000 of the solar energy Earth absorbs daily.
Nuclear plant + 500 MW data center
| Component | Power (MW) | Notes |
|---|---|---|
| Thermal input | ~550 | Fission heat |
| Waste heat at plant | ~50 | Cooling |
| Electricity to data center | 500 | Delivered output |
| Waste heat at data center | 500 | All electricity becomes heat |
| Total waste heat to environment | ~550 | Combined |
Nuclear is dramatically more efficient. Even the gas case is vanishingly small relative to daily solar absorption. Claims that treat a single facility as climatically decisive are orders of magnitude off the physical scale.
Technical detail and units examples:
MassEnergyUnits — data center power
·
Energy Abundance platform →
Nameplate Is Not Continuous Draw
Modern hyperscale and AI facilities are engineered so power and cooling track real demand:
- Servers use dynamic voltage and frequency scaling, idle/sleep states, and power-off of racks or floors when utilization is low.
- Cooling systems use variable-speed drives and zone control so they track real heat load.
- Power distribution is modular; sections can be de-energized without taking the whole site offline.
- Workload orchestration consolidates jobs so unused capacity can be powered down between large training runs or during low inference traffic.
The 500 MW figure is a design ceiling, not a constant. High capital cost of AI accelerators creates strong pressure to keep utilization high — but the engineering ability to scale down exists and is used.
PUE (Power Usage Effectiveness = total facility power ÷ IT equipment power) multiplies remaining IT load by facility overhead. Typical hyperscale values run about 1.1–1.4 under good conditions. At very low load, fixed plant can raise effective PUE, so facility draw does not fall quite as far as IT power alone. Even so, deep reductions remain large:
| Case | IT Power | Approx. PUE | Facility Draw |
|---|---|---|---|
| Peak | 500 MW | 1.2 | ~600 MW |
| Scaled | 50 MW | 1.5 | ~75 MW |
Waste heat at the site tracks facility electrical input. Nameplate capacity is a poor proxy for continuous environmental load.
Policy Honesty
If the real concern is energy and infrastructure scale, two coherent levers exist:
- Supply / abundance — expand generation and transmission (nuclear, hydro, gas where needed) and allow facilities so load can be met.
- Demand — price or restrict the end uses that create the load (heavy training, high-volume services).
What is incoherent is the third path often taken in practice: obstruct data-center siting while leaving AI demand politically unconstrained. Servers exist to serve load. Blocking local supply without reducing demand mainly exports the activity, delays planning, and creates scarcity. The joules do not disappear; they simply move.
Washington’s competitive position depends on choosing the first path. Energy abundance is the prerequisite for hosting high-value compute without scarcity pricing for everyone else. Restricting the physical plants while celebrating the applications that require them is the opposite of serious industrial strategy.
The Stakes
Places that build the infrastructure capture construction employment, long-term operating jobs, property and business tax base, and the surrounding ecosystem of suppliers and skilled talent. Places that block it watch those gains leave. The historical pattern is consistent across earlier industrial transitions. The same pattern is visible now with AI, automation, and energy-dense sources.
Seattle’s moratorium does not protect ratepayers or the climate in any meaningful way relative to the scale of the problem. It does signal to capital that Washington is a higher-friction jurisdiction for the defining infrastructure of this decade. Other states and countries are competing aggressively for exactly that capital.
This is the opposite of energy abundance. Blocking the facilities that create concentrated demand for firm power makes it harder — not easier — to justify and finance the nuclear, hydro, and gas resources our grid needs.
My Plan
Full-cost rates for large loads
Require data centers to pay the full marginal cost of service so residential and small-business rates are protected.
Standards, not freezes
Clear, technology-neutral rules for water efficiency, noise, and heat recovery — especially where district energy already exists — instead of freezing applications.
Expand firm generation
SMRs, preserved and modernized hydro, and natural gas as bridge/peaker capacity so the region can host high-value loads without scarcity pricing.
Skills and shared returns
Align workforce training with the complementary jobs these facilities create, and structure tax and ownership mechanisms so ordinary Washingtonians share in the returns.
Throttling the tools does none of these things. It only makes the adjustment steeper when the technology arrives anyway — and it will.
Washington can choose to build or it can choose to fall behind. There is no third option that keeps the prosperity while rejecting the infrastructure that produces it.
Innovation beats coercion. Abundance beats scarcity politics. Results beat ideology.
Together, we can make Washington golden.
— Jillian England
Candidate for Washington State Senate, Legislative District 36